From owner-freebsd-hackers Tue Jan 24 17:44:24 1995 Return-Path: hackers-owner Received: (from root@localhost) by freefall.cdrom.com (8.6.9/8.6.6) id RAA21508 for hackers-outgoing; Tue, 24 Jan 1995 17:44:24 -0800 Received: from gndrsh.aac.dev.com (gndrsh.aac.dev.com [198.145.92.241]) by freefall.cdrom.com (8.6.9/8.6.6) with ESMTP id RAA21490 for ; Tue, 24 Jan 1995 17:44:20 -0800 Received: (from rgrimes@localhost) by gndrsh.aac.dev.com (8.6.8/8.6.6) id BAA11440; Wed, 25 Jan 1995 01:43:31 GMT From: "Rodney W. Grimes" Message-Id: <199501250143.BAA11440@gndrsh.aac.dev.com> Subject: Re: disklabel (1.1.5.1), partitions To: julian@tfs.com (Julian Elischer) Date: Wed, 25 Jan 1995 01:43:31 +0000 (GMT) Cc: kuku@gilberto.physik.rwth-aachen.de, freebsd-hackers@freefall.cdrom.com In-Reply-To: from "Julian Elischer" at Jan 24, 95 11:07:44 am X-Mailer: ELM [version 2.4 PL23] Content-Type: text Content-Length: 1830 Sender: hackers-owner@FreeBSD.org Precedence: bulk > > > > > > > I had disk problems on 'gil' recently (hopefully solved) - reformatted > > a Fuji M2694S using SCSICNTL. > > > > I was wondering about the following when disklabeling again: > > > > dmesg told me: > > > > 2117025 1819 cyl 15 heads 77 sec > > > > Using these figures in a disktab entry resulted in a warning saying that > > partitions c,d,e exceeded end of disk. (I made them p[cde]#2117025). > > > > So I reduced them to what 15 * 77 * 1819 calculates to, namely 2100945 > > and disklabel kept silent. > > > > Did I give away some sectors ? > > > Unfortunatly yes... Infact you gave away 2117025 - ( 15 * 77 * 1819) = 16080 sectors, if you had used a translation of (64 heads * 32 sectors * 1033 cyl) you would have only given away 1441 sectors. > the question is "What do we do with disks with variable geometries.. > especially when the disklabel code assumes it is a fixed geometry. > > we calculate an 'average' number of setors per track.. > but it has to be an integer, and that truncation leads to a bunch of > unused sectors.. And this rounded integer is one of the worse things to use when creating a disklabel for the fact that you may give away up to (cyl * head) - 1 sectors instead of the max of translated head * sec when using translated or fictitios geometries. I highly recomened to anyone using the probe reported geometry to think twice and convert to a 32 sector 64 head geometry. > julian > > --Chris Christoph P. U. Kukulies kuku@gil.physik.rwth-aachen.de > > FreeBSD blues 2.1.0-Development FreeBSD 2.1.0-Development #1: > > Wed Jan 18 10:42:31 1995 kuku@blues:/usr/src/sys/compile/BLUES i386 > > > > -- Rod Grimes rgrimes@gndrsh.aac.dev.com Accurate Automation Company Custom computers for FreeBSD